Re: Precision in Performer...
Jim Helman (jimh++at++surreal)
Thu, 16 Dec 93 12:48:37 -0800
I belive you are simply exceeding the limits of single precision
floating point numbers. If your object is one unit large and you
position the object at (x, y+1, z) and the eye point at (x, y, z) with
x, y, z large enough so that in the FP representation (y+1) == y, then
the position relative to the eye point (y+1)-y is going to suffer
severe precision problems. Normally, single precision floating point
should get you to around 1.0e7 before suffering one unit jerkiness,
but it's plausible that intermediate calculations might cause a loss
of precision earlier.
One solution is to maintain double precision tile offsets, model the
tiles near the origin and handle the translations yourself so that
neither the DCS nor the ChanView translations get too large.
Angles substantially outside the range of -360 to 360 can also cause
problems, but you indicate your angles stay small.
rgds,
-jim helman
jimh++at++surreal.asd.sgi.com
415/390-1151
This archive was generated by hypermail 2.0b2
on Mon Aug 10 1998 - 17:50:07 PDT